Skip to main content
2026/9/24 16:47 (UTC+8) · New Model · xAI 🚀 grok-4.7 is live, on a new 2.1T base model at the same price as Grok 4.6 SpaceXAI released it on September 21. Parameters grow from 1.5T to 2.1T, and reinforcement learning puts more weight on multi-hour tasks. In independent testing, Terminal-Bench 4.0 rises from 18% to 33%, with the biggest gains in coding agents.
  • Pricing: $2 / $6 per million tokens, $0.50 for cached reads, second tier above 200K, matching the provider
  • Groups: default / svip / CodexResponses
  • Context: 500K, with text and image input
Note: the unit price is unchanged, but it uses about twice the output tokens per task as Grok 4.6, so compare a bill on real tasks before switching. Full write-up: Grok 4.7 Is Live.
← Back to Live Updates · 📚 Monthly archive